A Few Thoughts About Artificial General Intelligence

UBI, Enclosure, and the Owned World

Part One: The End of All Songs

I’ve spent nearly my entire adult life writing on shared electronic fora of one kind or another. I got a late start on literacy, but I took a typing class in high school, to fill my vocational education requirement. It was a revelation. From then on, anything with a keyboard was absolute catnip to me, and when the Internet came along all I could see was an opportunity to write for more people. Alas, I was training my own replacement. AI is now, or soon will be, as good or better a writer than I could ever be. That’s not an opinion. It’s a fact.

This isn’t just a problem for me, of course. The trillions of lines of computer code, assiduously composed by skilled technicians during decades spent in neon-lit veal pens, is exactly the same kind of fertile medium for AI that listservs, BBSs, Usenet, IRC, blogs, social media, stolen books, and online versions of various legacy media have proven to be. Interestingly, when I talk to various people about whether or how soon AI will eat their jobs, I’ve found computer programmers to be among the most resistant. Which seems odd, since they and their kind have made generations of secretaries, typists, and mailroom clerks obsolete. They’ve killed movie theaters and broadcast media (and, by extension, broadcast media receivers, like televisions and radios), newspapers, analog telephones, and the physically printed and mailed personal letter. Most programmers I’ve known have defended these obsolescences as the drumbeat of progress, shrugging philosophically about the layoffs and the bankruptcies, muttering about buggy whips. And yet. Many of them go a bit tharn at the suggestion that the service they provide to the world (that service being, after all, entirely digital, and text-based, and even less prone to colloquialism and onomatopoetical flights of fancy than my own field) could be adequately and cost-effectively provided by a machine. The fact that they helped build the machine in question doesn’t seem to improve their opinion of its capabilities. It always strikes me as an odd bit of sentimentality and bias coming from the original hard-nosed technocratic realists. Ah well. LessWrong, heal thyself.

I’m more concerned by the surprisingly universal lack of economic, psychological, or institutional preparation for the disaster that seems, to me, to be lurking just around the corner: a world in which we need many fewer people in order to do most of the things we do now. How many fewer? Nobody seems to want to say, and I honestly can’t blame them. The truth, even before it becomes a reality, may be more than anyone’s ready to take on. I think “half” is a conservative estimate based on some current structures and trends. If someone said a tenth as many workers overall, I wouldn’t dismiss that number out of hand. I think the limiting factor won’t be the technology, but rather how quickly companies can eliminate human labor while incurring acceptable losses in terms of political, social, and economic backlash. If someone invented cheap artificial general intelligence tomorrow, I think it would still take years, at least, to be fully utilized. The lag wouldn’t be technological or practical. It would be corporations and governments trying to figure out how to implement the new model without being murdered and having their factories and computers burned to the ground. One assumes the AGI itself would be able to provide the captains of industry with both individualized and universal surveillance, and analysis of the data generated by that surveillance. The analysis and tactical suggestions the AGI produced would undoubtedly be better-informed, more pragmatic, and more subtly effective than whatever the sentimental and biased human executives who “own” it might come up with.

Though one mustn’t be too hard on sentiment and bias. I think the circle of sentimentality and bias surrounding a hard locus of ownership will soon be the only thing keeping anyone employed. The entertainment industry already works this way. The world has enough talented artists to fill the ranks of Hollywood a thousand times over. Since everyone (not everyone, obviously, but enough people to create a vast, undifferentiated pool of applicants) is equally qualified, the only distinguishing characteristic becomes access. When AGIs mean that a person with a third grade education can “run” a Fortune 500 company as well as someone with a PhD, an MBA, and 20 years of experience, the only meaningful criteria for deciding who gets to occupy that position will boil down to access. Sentiment. Bias. Inherited wealth, and connections. It’s already true that something upward of 70% of jobs are won through networking. Do we think that’s going to be more true or less true in the future?

Every so often someone produces a list of jobs they think are going to last through AI. I’m usually struck by how many of those jobs are social welfare jobs, like “teacher” “therapist” and “social worker.” It all seems very unlikely to me. When only a small fraction of people have actual income, and they fight taxes every inch of the way, we’ll only be able to fund a small number of social welfare workers. For everyone else it’ll be pure entrepreneurship in the non-social-welfare sectors where AI doesn’t dominate. Do we really think access to jobs will be more merit-based and less nepotistic when the only private sector work left is some variation of “childcare worker for parenting rich people,” “nurse for dying rich people,” “sex worker for horny rich people,” and “vice president nominally-but-not-really in charge of the part of the AI that deals with quality control”?

Everybody else is going to be surplus to requirements. They won’t be herded into camps or anything. Most of them won’t even be properly fired. (Nothing so narratively provocative.) But they (we) may be, I suspect, forced to fade away. Daycare in coastal cities runs upward of $20k a year. Elder care is basically a giant siphon for vacuuming up any personal wealth a family might somehow have hung onto. Housing is getting more expensive at some appalling multiple of background inflation. Cut back on health insurance, leave the lead in the drinking water, don’t trouble yourself about air quality, and let the factory farms dust the crops with whatever they want… well. The situation won’t exactly take care of itself. Not entirely. But in a generation or two you’ll have a lot fewer people begging you for change on your way from your robot car to your robot hotel. Make people miserable and scared enough, make sure their only escape is streaming media and video games, they won’t even meet each other to screw, let alone reproduce. More and more these days I have the uncomfortable feeling that some of the weirdest science fiction I read when I was in high school might also turn out to be the most accurately predictive.

Near-term, a lot of the wonks seem to think that AI will lead to a Universal Basic Income. I don’t know what to make of that one either, given our culture’s decidedly mixed attitudes toward people marginalized by systemic unemployment. We’re having a debate right now (again) about whether we should bring back coal mining, which hits the trifecta of being unhealthy, unnecessary, and unprofitable. Republicans seem to believe that hiring people to stand around punching themselves in the face for ten hours a day, with an unpaid mandatory one-hour lunch break, is still somehow both better for the economy and more ethical than just handing them the pittance they’d earn in the mines and letting them write Appalachian poetry, or sit around in their back yards shooting cans with BB guns or whatever. I have no trouble imagining a future where literally every job in the world is done by AI, and I’ll still have to watch some Republican congressman with a silver crew-cut and a throbbing forehead vein yell about how 70% of able-bodied adults just don’t want to work because a diamond mine in Alabama is offering people $1 an hour to polish rocks by swishing them around in their mouths.

This is where I start bumping up against the outer edge of the Overton window. Because the only honest way to talk about solutions to all this is to start talking about “the system” and a bunch of history that’s different enough from most people’s understanding that it ends up sounding like I made it up.

I’ll give it a shot anyway.

Part Two: A (Genuinely) Brief History of Poor People Getting Screwed With Their Pants On

Between the 17th and early 19th Century, the United Kingdom and a few other places went through a process of what we now call enclosure. Before the enclosure movement, most land in the U.K. wasn’t actually privately owned in the way we think of it today. Peasants were both owners of the land they lived on, and property. They weren’t allowed to leave, they were required to pay taxes — but there was no legal mechanism for forcing them off the land. In the 1600s, various forces conspired to expand and strengthen the international market for British textiles, at roughly the same time that something called the British agricultural revolution — centralized irrigation, crop rotation, improved plowing techniques, draining of swamps, etc — started to come into the picture. The rising merchant class wanted access to arable land for factory farming, so they invented a legal mechanism called enclosure that allowed them to steal it. The basic trick was to get signatures from a certain percentage of the peasants who lived on the land, supposedly selling it to the merchants, usually for some amount of money vastly below what the land was worth. Then even the peasants who didn’t sign could be removed, and the land would be turned into, essentially, a plantation. Obviously there was a lot of fraud, and sometimes violence, involved in this whole process.

Please note that, in the United States, this package of legal, technological, and social tools is described in many of our high school history textbooks as the “European” paradigm that was utterly alien to Native Americans. The sentence in the book usually goes something like this: “Native Americans often signed treaties turning land over to European colonists because the idea of private land ownership was foreign to them. They may not have fully understood the meaning of the treaties.” The reality is that, at the time these treaties were being used to rob Native Americans of their land, European merchants and lawyers were using the exact same techniques to rob European (and especially British) peasantry. Lack of understanding wasn’t the problem. The Anglo-American courts of the time were trained to ratify the thefts with or without the participation of the people being robbed. And the paradigm of private land ownership wasn’t “European” in the 17th Century anymore than private tiger ownership is “American.” It was a thing done by rich assholes who were European — but it wasn’t a thing Europeans in general did. In fact, many of the poor white colonists being used as shock troops for the invasion of North America were first- or second-generation refugees of European land-theft, for whom the idea of “private land ownership” only had currency because it had just been used to kick their family off a farm they’d lived on for 500 years.

The other thing that maybe doesn’t get talked about as much as it should is that kicking people off their land served two functions for the rising merchant class in Europe. It consolidated access to raw materials — but it also turned the former peasant farmers into unwilling fodder for the manufacture of trade goods, even before what we now call the Industrial Revolution. This has been the formula for the success of capitalism since its inception: invent legal mechanisms to deprive people of resources that belong to everyone, make the dispossessed work to pay for access to the goods made from what was once theirs, pretend it was all a choice. This formula has nearly always resulted in some form of surplus population. Up to a certain point, the ownership class gets more benefit from having too many workers chasing too few jobs than they do from the opposite state of affairs. When the “certain point” is reached and exceeded, the pet philosophers of the ruling class come up with ideas like the Malthusian theory, which declared that the only humane and rational way to keep the population of poor people from swamping the country was to create bureaucratic mechanisms that would intentionally but impersonally starve them. Limit the number of kids they could have by limiting the amount of food they could get. If they had more kids than they could feed, the kids would starve and the family would shrink back to its administratively dictated ideal size. It’s sort of a eugenics version of the price feedback mechanism. Modern conservative attitudes on welfare are still deeply rooted in the Malthusian worldview.

Part Three: So…?

Artificial intelligence is the new enclosure movement. Private corporations have trained their AIs on the digitized archive of all human knowledge (including stolen copies of my book). If there’s a resource that can be said to be more the common property of all humanity than the content of the World Wide Web, I’m not sure what it might be. The owners of AI have exploited that resource for themselves, and lay claim to the product. Now the wealthiest corporations in the history of the world propose to use the machines they’ve built from our collective knowledge to replace most of us in the workplace, and leave the refugees to find what shelter they may in the labor “market.” Within the dictates of our current model of capitalism, the result will be the same as it was last time: we’ll be cut loose from the old economy, left homeless, desperate, and economically irrelevant. We’ll be told it’s all our fault, because capitalism wastes nothing. If we can’t earn enough to survive, it must be because we aren’t trying hard enough. The wealthy will employ as many of us as they have a use for, and resort to some updated and subtler form of market-based Malthusean population control to keep the rest of us from breeding — or at least make sure our kids create as few hassles as possible.

None of that requires a change in our current system. We already relegate people to obsolescence on a scale that would have seemed insane just a few decades ago. And to those hoping a UBI might save us without significantly modifying the way our economy works, I’d point to (among other things) what Trump is currently doing to his political enemies. Now imagine you’re sitting in your tiny state-subsidized apartment, in some exurban backwater with no market competition for affordable housing, collecting your UBI, watching some reality TV nightmare on streaming — there are no jobs to be had, no Post Office, no public parks — and you’re told that of course we can’t keep giving the UBI to people with felony records. Then we can’t keep giving the UBI to people with drug problems. Then we can’t keep giving the UBI to people who write subversive things on the Internet. Then we can’t keep giving the UBI to people who don’t go to the right church. In a world where everything is owned, blocking people from buying — from being able to have money, and participate in the market — is essentially a death sentence. We can say we’re cutting people off from the privilege of a UBI. But what we’re really doing is depriving them of any legal means of existing. Is that hard to imagine? Given what’s happening right now? I don’t think it is. So that’s basically where we’re heading. In fact, we’re already several long strides down that road.

And here, again, is the Overton window. Because the next thing I need to say is, “Unrestrained capitalism has taken us about as far as it can.” And that’s what fuzzy-headed radicals say. It’s crazy talk. Capitalism is eternal. It’s the best system. We can prove it, with some scratch paper, and a pen, and an etch-a-sketch.

Except of course that’s bullshit. Pure capitalism has never existed. Controls on monopoly dilute the purity of capitalism. Controls on drugs and alcohol, food safety laws, the whole legal concept of torts; all are impurities, turning the brittle metal of capitalism into some more humane alloy. Not allowing individuals to sell themselves into slavery under the principle of freedom of contract is an adulteration of capitalism’s purity. Which is fine, because pure capitalism is a childish myth. It’s a fairy tale, a religious concept rather than something that can or should exist in the world. So we should let that go, and move on to the problems at hand.

Part Four: An Old Sicilian Proverb

Markets are one tool out of many. They’re generally thought to be the most efficient way to allocate scarce resources. But they aren’t awesome at not making resources scarce in the first place. Ask the dodo bird. Or the passenger pigeon. Or, you know, in a couple of years, the honey bee. Ask the rhinos, with the horns people think make good boner medicine. That’s not even a real thing, and it’s destroying the resource it’s based on. I guess there’s an argument to be made that if we could convinced a billion people that sprinkling microplastics from the middle of the Pacific on their dicks would give them a 24-hour hard-on, we could make the market conserve something for a change. But somehow it never seems to work out that way.

Some economists have modeled systems where markets are used only and exclusively in areas where we want competition to breed efficiency. If we could stop all poachers, and feed the rhino-horn market with horns ethically harvested from rhinos that died of natural causes, market competition would maximize profits and make transport and delivery as efficient as possible. Market competition between bricklayers could produce efficiencies and improve outcomes, and so on. But there are many areas where markets are actively harmful.

Economist Karl Polanyi described a few such areas when he wrote about the fictitious commodities of land, labor, and money. The idea behind land, labor (meaning not just work, but the people who do it), and money being “fictitious” commodities, is that no sane person would want the market to actually be able to generate a value of “zero” or “infinity” for any of those things. Trading on a market seems like as good a way as any for deciding how people should be compensated for their labor, until you imagine someone whose labor is regarded as having “zero” value. To go back to enclosure, the earth was taken from all humanity, and handed over to the markets. We allow it because, supposedly, that system benefits everyone. But the person whose labor is regarded as having “zero” value (or even just a value insufficient for survival) finds themselves with no option for self-help. They live in an owned world, with no market for the only thing they have to offer in trade. Real estate markets create a similar problem. If all land can be owned, then one person can own all the land; everyone else’s balance can be “zero,” and the value of the resource is effectively set to “infinity.” So once again, those with a balance of “zero” have no recourse for self-help. And so on, with money. That these sorts of things are never supposed to happen doesn’t alleviate the existential threat implicit in the fact that they could happen (and do happen, sometimes, in some limited contexts). Personally I would be inclined to add potable water, breathable air, and healthcare to the fictitious commodities — and there are plenty of real-world examples of people being cut off from legal access to those things. Treating the basic elements of life like trade goods produces perverse and morally indefensible results.

We’ve been trained to respond to this kind of attack on capitalism by throwing up our hands and asking what the alternative is. The answer isn’t very mysterious. Most of us could probably even come up with it, if we thought about it for a bit. Declaring a commodity ineligible for exchange on the market doesn’t mean it gets handed out for free. It means we make decisions about it that aren’t predicated primarily on efficiency, or the lowest bid. And if that seems utopian, and too far from our current model to even be contemplated, I’d point that that the United States has used the military as an employer of last resort for nearly a century. Even now, some of the people who can’t find what they need in the market can join the armed forces and get clothes, shelter, food, education, work, and medical care. Those are things that should be available to everyone, without having to sign up to be killed, or kill strangers, on the order of someone like Donald Trump. And we have done a peacetime version of national service before: in the years prior World War II, the United States responded to persistent systemic unemployment with the Works Progress Administration, the Civilian Conservation Corps, the Federal Theater Project, the Federal Art Project, and so on.

I don’t know what’s best for us in the face of AI, and the possibility of AGI. I like to imagine a rebirth of the New Deal programs. I like to imagine a shift in societal priorities. But I also know that the solution for new problems is almost never behind us. What I can say from here is that the economic and cultural model we’re currently using has a demonstrated history of causing widespread misery and expropriation during paradigmatic shifts in modes of production. Other models have been theorized, and many have been successfully implemented both here and abroad.

Everything else is just inspiration, ideas, and arguments.

But no shit: we need to figure something else out, or we’re fucked. History does tell us that much.